Re: [rtc-linux] [PATCH 5/6] rtc: max77620: add support for max77620/max20024 RTC driver

[Krzysztof Kozlowski]

On 12.01.2016 02:07, Laxman Dewangan wrote:
> 
> On Monday 11 January 2016 09:34 PM, Alexandre Belloni wrote:
>> On 11/01/2016 at 18:47:34 +0530, Laxman Dewangan wrote :
>>> On Friday 08 January 2016 07:06 PM, Laxman Dewangan wrote:
>>>> On Friday 08 January 2016 07:06 PM, Mark Brown wrote:
>>>>> * PGP Signed by an unknown key
>>>>>
>>>>> On Fri, Jan 08, 2016 at 06:34:29PM +0530, Laxman Dewangan wrote:
>>>>>
>>>>>> If we get the parent device, regmap handle and interrupt number from
>>>>>> mfd
>>>>>> core independent of the PMIC (MAX77620 or MAX77686), then same driver
>>>>>> can be
>>>>>> used here.
>>>>>> Two way which I can think of here:
>>>>> Parent device is just dev->parent, you can use dev_get_regmap() to
>>>>> get a
>>>>> regmap given a struct device and you can use platform resources to
>>>>> pass
>>>>> the interrupts to the children from the MFD (there's some examples,
>>>>> wm831x is one).
>>>>>
>>>>>
>>>> I think it should work with named regmap. mfd whould init regmap
>>>> with name
>>>> and rtc driver should ask with same name.
>>>>
>>>> I saw three drivers which looks same:
>>>> rtc-max77620.c (new from me) and already available rtc-max77686.c,
>>>> rtc-max77802.c
>>>>
>>>> Seems I can develop IP based rtc driver as rtc-max77xxx.c
>>> I came with one of issue when doing this.
>>>
>>> The RTC driver parent is not the same parent for which i2c slave
>>> address get
>>> registered.
>>> There is two slave address from max77620, 0x3C (for general) and 0x68
>>> for
>>> RTC.
>>>
>>> In max77620 mfd driver, we make dummy i2c client for 0x68 and initialize
>>> regmap with this address.
>>>
>>> Now on mfd_add_devices, we pass the device for 0x3c and hence the RTC
>>> driver
>>> treat the parent as the 0x3c device but actually it should be 0x68 to
>>> get
>>> the proper regmap.
>>>
>>>
>>> Two approach:
>>> 1. If we add the option to pass parent_dev when adding cells form
>>> mfd_add_devices and select the parent device based on this option
>>> then it
>>> can be easily handle.
>>>      Add parent_dev structure in struct mfd_cell and then change the
>>> parent
>>> in mfd_add_device() if cells has parent device.
>>>
>>> 2. Register the RTC driver with different mfd_add_devices with dummy i2c
>>> client device.
>>> So two times mfd_add_devices.

Lexman,

I don't quite get the problem. This looks exactly the same as for
max77686. What is the difference? I don't see any need to change the
mfd_cell for current drivers...


>>>
>>>
>>> IMO, approach 1 looks good to me.
>>>
>>> Any opinion?
>>>
>> If the RTC is not at the same address, I'd say this is not an mfd
>> anymore, can't you probe it directly from DT?
>>
>>
> This approach is also possible but,
> 
> although this is independent IP with separate i2c address but became it
> is inside the PMIC and its interrupt depends on PMIC internals, I like
> to register this rtc device from the mfd core.
> So that when mfd core is ready with their interrupts and initial
> setting, it can register rtc device.

Alexandre,

All of these devices have some of the blocks under separate I2C address.
It is still a MFD because for example interrupts are shared and governed
by parent.

Best regards,
Krzysztof